TSMC victory lap: 10nm tape-out in Q1, 7nm HVM in H1'18

witeken

Diamond Member
Dec 25, 2013
3,868
11
106
#1
http://seekingalpha.com/article/396...-results-earnings-call-transcript?part=single

Our 7-nanometer technology, N7, the technology development is well on track. N7 is a further extension of N10 technology with more than 60% in logic density gain and 30% to 40% reduction in power consumption. N7 fully leverages N10 e-learning and shares more than 95% of common tools.

We have expanded our N7 design ecosystem development to include both mobile and high-performance computing, to enable our customers to deliver their first-to-market products. Our N7 adoption is very strong with customers ranging from mobile GPU, game console, FPGA, network processors and other consumer product applications. We have more than 20 customers in intensive design engagement with us and expect to have 15 customer tape-outs in 2017. The volume production of N7 will start from first half 2018.
 

Burpo

Diamond Member
Sep 10, 2013
4,063
33
126
#2
Victory lap? Ok..

"16 and 20 nanometer contributed 23% of our total wafer revenue in the first quarter. In addition, 28-nanometer saw a nice rebound in demand and contributed 30% of our total wafer revenue. Together, these three advanced technologies accounted for 53% of total wafer revenue."
 
Nov 4, 2012
156
0
81
#3
Victory lap? Ok..

"16 and 20 nanometer contributed 23% of our total wafer revenue in the first quarter. In addition, 28-nanometer saw a nice rebound in demand and contributed 30% of our total wafer revenue. Together, these three advanced technologies accounted for 53% of total wafer revenue."
This advancements gets them Apple contracts. Apples contracts are mostly what allows them to so aggressively invest in more advancements (one of the very few willing to pay the premium).
 

Exophase

Diamond Member
Apr 19, 2012
4,440
0
81
#4
That usage of "e-learning" threw me off. I thought they were talking about some kind of adaptive technology but it just means they're providing online seminars to their customers.
 
Mar 10, 2006
11,719
122
126
#5
That usage of "e-learning" threw me off. I thought they were talking about some kind of adaptive technology but it just means they're providing online seminars to their customers.
Faulty transcription. "yield learning" is what the exec said.
 
Jan 23, 2015
177
0
36
#9
On a related note
http://www.anandtech.com/show/10272/samsung-foundry-updates-7nm-euv-10lpp-and-14lpc


From the transcript:
Low: Will EUV be ready at 7nm?
B. Suh: We are reviewing the possibility of EUV adoption very carefully and readiness for mass production will be determined accordingly.
http://www.samsungsemiblog.com/foun...g-excellence-and-advanced-technology-updates/

Overall it seems somewhat vague.Still there's no indicator how well EUV be ready for primetime by then.
 

witeken

Diamond Member
Dec 25, 2013
3,868
11
106
#10
Huge blow to SS and Intel: 10nm got in fact taped out in 2015 (December), not in Q1 as was reported earlier (although one can argue if tape out should mean silicon back from fab or put in).

http://anandtech.com/show/10329/arm-announces-10ff-artemis-test-chip

One could also argue that 0.7x power *for a low-power chip* is not too good, a bit like 20nm.
 

sm625

Diamond Member
May 6, 2011
8,176
1
106
#11
Anyone want to bet that we wont see a retail phone with a TSMC 10nm part until fall 2018 at the earliest?
 

krumme

Diamond Member
Oct 9, 2009
5,786
172
136
#12
Huge blow to SS and Intel: 10nm got in fact taped out in 2015 (December), not in Q1 as was reported earlier (although one can argue if tape out should mean silicon back from fab or put in).

http://anandtech.com/show/10329/arm-announces-10ff-artemis-test-chip

One could also argue that 0.7x power *for a low-power chip* is not too good, a bit like 20nm.
Mobile for sure is still on a roll vs pc market with all that hefty investment in new process ! The perf and leakage is meh...but density is very promising. Up to 2.1 vs 16nm as Andrei said. Leakage took a nice dive with finfets now its density.

Interesting stuff all over. The rush for 7nm so quick. And the arch for Ares. Perhaps some beefy fpu comming up?

Man its so few years ago arm11 was the thing...damn.
 

krumme

Diamond Member
Oct 9, 2009
5,786
172
136
#13
Anyone want to bet that we wont see a retail phone with a TSMC 10nm part until fall 2018 at the earliest?
I take that bet.
Lets call it the apple bet :)

Edit: you might be right but then they have been stacking tens of millions since h1 2018. But i still take it non the less. Ares on 10nm h1 2018 in a retail phone.
 
Last edited:

witeken

Diamond Member
Dec 25, 2013
3,868
11
106
#14
10nm end of year according to ARM according to EETimes. Should be competitive with Intel 14nn, but Intel 14nn will have been on market for 2 years, so nothing to be ashamed of for Intel.

http://www.eetimes.com/document.asp

Edit: wait, don't see how that can be right, given that 10nm has only just taped out.
 
Last edited:

Andrei.

Senior member
Jan 26, 2015
269
32
116
#15
Anyone want to bet that we wont see a retail phone with a TSMC 10nm part until fall 2018 at the earliest?
Pretty sure we'll have devices in 2017 from more than one vendor.
 
Mar 10, 2006
11,719
122
126
#18
10nm end of year according to ARM according to EETimes. Should be competitive with Intel 14nn, but Intel 14nn will have been on market for 2 years, so nothing to be ashamed of for Intel.

http://www.eetimes.com/document.asp

Edit: wait, don't see how that can be right, given that 10nm has only just taped out.
Why bring up Intel? They are 100% irrelevant right now in the areas that ARM is talking about.

Also, I doubt that Intel 14nm electrical characteristics are as good as TSMC (and probably Samsung) 10nm.

Intel has no process lead.
 

Abwx

Diamond Member
Apr 2, 2011
9,006
374
126
#19
.but density is very promising. Up to 2.1 vs 16nm as Andrei said. Leakage took a nice dive with finfets now its density.
60% better density is 1.6x improvement, or eventualy 0.625x the area, or 37.5% area reduction...
 

witeken

Diamond Member
Dec 25, 2013
3,868
11
106
#20
Why bring up Intel? They are 100% irrelevant right now in the areas that ARM is talking about.

Also, I doubt that Intel 14nm electrical characteristics are as good as TSMC (and probably Samsung) 10nm.

Intel has no process lead.
Okay, so that's why ARM says 10nm will offer literally zero performance improvement.

Will it have air gaps? How high will the fins be?

LOLOL.
 
Mar 10, 2006
11,719
122
126
#21
Okay, so that's why ARM says 10nm will offer literally zero performance improvement.

Will it have air gaps? How high will the fins be?

LOLOL.
I don't see anything that says 10nm won't enable a significant performance improvement.

No idea if it'll have air gaps. 16FF+ had 42nm tall fins, just like Intel 14nm, so I would expect them to be taller than Intel's in TSMC 10nm.
 
Oct 27, 2014
172
0
46
#22
Did they had a stock fall lately and they wanted to pump it up? :D
So many missed promised these guys...
 
Dec 17, 2008
1,850
7
91
#23
I personally do not care what you name the fabrication processes with terms like X nanometer for you can't compare different companies process even if they claim to be the same or similar scale of process.

One 28nm is not like another 28nm and so on. TSMC has 5 different 28nm process under the same roof for example.

low performance (LP)
high k metal gates (HP)
low power with high-k metal gates (HPL)
high performance mobile computing (HPM)
high performance compact mobile computing (HPC)

and all of those 5 different processes have different characteristics even though scale wise they are very similar. Here is the TMSC feature comparison chart in response to voltage

http://www.tsmc.com/english/dedicatedFoundry/technology/28nm.htm

------

Thus the only thing I care about is its performance characteristics. Now performance can be compared to many things, like voltage, idle power, average power, total performance, die area for various components, and so on.

And even if the fabs process has a great performance you can fab a chip in multiple different ways and make different trade offs such as create a cortex A chip but design it to work at only lower voltages and lower performance but has great power characteristics or design it for total maximum performance that process allows you to do so

------

Thus all this info tells me is TSMC thinks their R&D on the process to actually delivering that process to customers is on track and is doing great. I hope so since TSMC is such a leader in semiconductors overall and especially on mobile semiconductors which I see as a more important field than some other specialities.

------

PS you should not compare the naming of Intel scale vs TSMC for TSMC and also Samsung has inflated their naming Scales recently where Intel has better density and power characteristics for a similar theoretical chip.

But once again the name does not matter but instead the properties of the chip. The name is just marketing bullshit.

Oh and since Intel is not making any mobile chips in the near future (near in my mind being the next 4 years), and it looks like TSMC and Samsung are going to be the big two foundries for mobile chips in the near future comparing a theoretical intel x86 mobile chip on a theoretical intel process vs a theoretical arm chip on a theoretical TSMC process is silly.

Instead compare a specific arm TSMC chip vs other arm TSMC chips and compare intel chips vs its older intel chips since the companies are no longer competing in mobile since intel has abandoned mobile in any sub 5 watt devices.
 

Abwx

Diamond Member
Apr 2, 2011
9,006
374
126
#24
PS you should not compare the naming of Intel scale vs TSMC for TSMC and also Samsung has inflated their naming Scales recently where Intel has better density and power characteristics for a similar theoretical chip.
Could you point us a "similar" chip where Intel has better density, sorry but node after node Intel has better density only when they have a node advantage, otherwise all is see are slides that are contradicted by real products transistors count, for instance, what is the density of SKL..?..

Other than this it makes no doubt that Intel s 14nm is better electricaly speaking than TSMC s 16FF+ but from early numbers published by GF this latter (and Samsung of course) have the best electrical chracteristics..
 


ASK THE COMMUNITY

TRENDING THREADS